When manufacturing high-resolution miniaturized camera modules, the step of integrating an objective comprises performing an active alignment process, i.e. actively aligning the objective (object lens) in relation to the image sensor while observing and evaluating the image that arises. This involves moving the objective in relation to the image sensor and evaluating the arising image in accordance with predefined quality criteria of the image sharpness (typically measuring the image contrast and/or the module transfer function [in brief: MTRF] at different positions in the image). Positioning is optimized, e.g., by maximizing the quality criteria measured, and the objective is fixated (e.g. by means of gluing) accordingly in this position in relation to the image sensor. One precondition necessitated for this is that the properties of the objective (e.g. image contrast, MTF) that are drawn on for the quality criteria will change to a sufficiently measurable degree over the shifts in position used in the process, as is known, for example, from US 2013/0047396 A1 or JP 20070269879.
As far as active alignment is concerned, conventional optimization algorithms will fail if the parameters of the objectives vary only slightly in relation to the positioning steps. The latter applies, e.g., to objectives having large depths of focus (and, in particular, multi-aperture objectives comprised of microlens arrays) wherein a change in the z distance between the objective and the image sensor results in only slight changes in the image sharpness that in real cases are difficult to measure.
Due to the rotationally symmetric layout of conventional objectives of miniaturized cameras about the optical (z) axis, industrial automatic assembly machines in most cases exhibit five degrees of freedom (and, accordingly, five axes) for relative positioning of the optics in relation to the image sensor (3x translation along x,y,z axes+2x tilting [tx,ty] about the x and y axes, as is depicted in FIG. 18, for example). Thus, the established active assembly processes and machines are not suitable of aligning objectives which comprise no rotational symmetry about the z axis. These include, for example, anamorphic objectives, objectives comprising directionally selective filter components, but also multi-aperture objectives consisting of microlens arrays.
FIG. 18 shows a schematic image of an assembly layout of multi-aperture imaging optics 12 to form an image sensor chip 16 with a notation of the necessitated degrees of freedom x,y,z (translation), tx, ty, and tz (rotation).
Both restrictions described apply in combination for multi-aperture imaging objectives, in brief multi-aperture optics, such as so-called electronic cluster eyes as are known from WO 2011/045324 A2. The multi-aperture arrangement consists of an array of optical channels that is one- or two-dimensionally extended in the x-y plane, each optical channel capturing a defined part of the entire object field.
The location of the central position of the aperture of each individual optical channel in relation to the center of the associated subimage (as viewed in the x-y plane in each case) here plays a major part in terms of accuracy of the reconstruction and/or the resolution capability of the overall image. The difference between the central position of the aperture and the central position of the associated subimage (pitch difference) has to be adjusted along the translational degrees of freedom in x,y with an accuracy of between half and one pixel pitch of the image sensor used.
This arrangement of multi-aperture optics has been developed specifically in order to realize miniaturized camera modules, in particular those having ultra-thin structural forms (for the purpose of being used in thin devices such as smartphones, tablets, laptops, etc., for example).
Accordingly, microlenses are employed therein which have very small focal lengths (e.g. f=1.2 mm) and, thus, large depths of focus. In accordance with the formula dz=4*/*(F/#)^2 for the depth of focus in the image space (dz) with diffraction-limited imaging of the wavelength W, a value of dz=12.7 μm for light of the wavelength of 550 nm and an f-number of F/#=2.4 is achieved, for example.
FIG. 19 schematically illustrates the requirements in terms of the alignment of multi-aperture optics 12 with an image plane BE of the image sensor 16. The multi-aperture optics 12 comprise several optical channels arranged in a one-dimensional or two-dimensional array and comprising a center. Optical channels located outside the center are configured to receive an obliquely incident principle ray PR. One recognizes that with oblique incidence, at the angle alpha “α”, of light of the principle ray of the central field point within an external optical channel, the point of intersection with the focal position (=e.g. temporary location of the image sensor during assembly) undergoes a lateral offset (“Δd”) within the depth of focus due to the difference of the z position (“Δz”). With a pixel pitch of p_px=2 μm of the image sensor and given the correspondingly large maximum lateral offset, the value of “Δz” is allowed to be Δz=4.3 μm at the most in accordance with the geometric relation tan(α)=Δd/Δz at an angle of incidence of α=25°. This value lies within the range of depths of focus, so that any existing active assembly techniques based on evaluating the image contrast do not allow sufficient accuracy in aligning the objective in relation to the image sensor when applied to multi-aperture imaging optics. Thus, FIG. 19 shows a schematic section through a multi-aperture imaging objective in accordance with WO 2011/045324 A2. What is shown are the principle rays for the average lines of vision of the optical channels. The magnification shows the lateral offset Δd of the center of the subimage of an external optical channel due to different focal positions Δz within the image-side range of depths of focus and of the angle of incidence a of the principle ray PR.
To illustrate this, a numerical example shall be given below.
Camera parameters comprise, e.g., a focal length (f) of 1.2 mm, a pixel pitch (ppx) of 2 μm, a range of vision having an angle of aperture of horizontally 59°, vertically 46° (diagonally 0°). A maximum angle of incidence (α) on the image plane amounts to 25°. Dimensions of the microlens array amount to (H×W): 7.75 mm×4.65 mm.
This results in associated alignment tolerances as follows: A tolerable shift in the x-y plane amounts to a maximum of 2 pixels, i.e. Δx≤4 μm and Δy≤4 μm. A tolerable twist about the x,y axis (wedge error) amounts to a maximum of half a pixel, i.e.
      Δ    ⁢                  ⁢          t      x        =                    arctan        ⁡                  (                                    p              px                                      2              ⁢                                                          ⁢              f                                )                    ≤              0.05        ⁢        °        ⁢                                  ⁢        and        ⁢                                  ⁢        Δ        ⁢                                  ⁢                  t          y                      =                  arctan        ⁡                  (                                    p              px                                      2              ⁢                                                          ⁢              f                                )                    ≤              0.05        ⁢                  °          .                    A tolerable twist by the z axis amounts to a maximum of one pixel in the external channels, i.e.
      Δ    ⁢                  ⁢          t      2        =            arctan      ⁡              (                              p            px                                H            /            2                          )              ≤          0.03      ⁢              °        .            A shift in the z axis (distance error) amounts to a maximum of one pixel pitch (Δd) in external optical channels->
      Δ    ⁢                  ⁢    z    =                    Δ        ⁢                                  ⁢        d                    tan        ⁡                  (          α          )                      ≤          4.3      ⁢                          ⁢              µm        .            
Known methods for aligning optics with an image sensor are known, for example, as active alignment and attempt to adjust individual lenses or entire assemblies in relation to an image sensor as a function of the quality (in most cases, of the contrast) of the respective image taken.
Known devices for active camera objective alignment primarily relate to assembling rotationally symmetric optics, so-called 5D active alignment, in relation to an image sensor in a production environment and for large numbers of items. Such devices and assembly techniques used are not modifiable to match the needs of active assembly of multi-aperture objectives. For example, an accuracy of the assembled axes is too small. For example, [1] describes that an x,y,z translation can be adjusted with an accuracy of ±5 μm, and a tx,ty, and/or tz twist can be adjusted with an accuracy of ±0.1°, which is insufficient for multi-aperture optics according to the above numerical example. The insufficient accuracy of the assembly processes is based on an evaluation of the image contrast, on a closed system environment and, accordingly, on a lack of access to driving the positioning system and to readout of the camera boards used. For example, a manufacturer of the device will specify the same test pattern, irrespective of which client (optics manufacturer) uses the device.
An assembly system which uses a combination of passive and active alignment is known from US 2013/0047396. Said system exhibits the same limitations as were described above.
A method of active camera optics assembly of several camera modules while using the evaluation of the image contrast is known from JP 20070269879. This method, too, is difficult or even impossible to adapt to the requirements of multi-aperture optics.
Alternative concepts describe an active objective holder. As an alternative to active alignment and fixation, imaging objectives may be mounted in holders that enable later positioning between the objective and the image sensor to be effected at a later point in time, as are described in US 2011/0298968 A1, for example. Additional feedback to the image sensor, an evaluation unit or a sensor is enabled by an active function such as autofocus or optical image stabilization. The designs necessitated for this involve a large amount of effort, are therefore costly and limit miniaturization of the camera modules. In the field of miniaturized multi-aperture optics or extremely miniaturized multi-aperture cameras, utilization of such micromechanical components is hitherto unknown for reasons of cost and in terms of reducing the size of the design.
Therefore, what is desirable is a concept enabling production of multi-aperture camera devices which comprise increased image quality and smaller production tolerances.